Context maintenance or switching is an important decision to be made by any conversation system. For example, for a query “show me action movies” followed by another query “with Tom Cruise,” the conversation system is expected to maintain context across the queries. But if the second query was “how about some comedy,” then the conversation system is expected to switch context. The conventional approach to solve this problem is to have a set of rules that determine whether the subsequent query is connected to the first query. However, rule-based systems are rigid and need programmers to be involved to address every possible situation that may arise during a natural language conversation.